Teesside University - Computing Project
What if you could train your own agents to perform intelligent actions in your game? What if your agents could learn from mistakes and improve on their own?
This project is a perfect example to show this. I developed a game simulator in Unreal Engine 4, where an AI agent has to solve a 2D platformer level by using his own senses and self-learning by experience.
The agent is managed by a neural network, which tells the agent what actions to perform depending on de surrounding environment and the nearest threats.